DMA for triton again...

DMA for triton again...

[Guennadi Liakhovetski]

Hello everybody

I know this subject has been discussed multiple times already, I've read a lot of posts about it, but still haven't been able to fix my particular case. So, here we go:

computer: Chipset 430FX / Triton / PIIX, disk Western Digital Caviar AC21600H Firmware code F6 (no UDMA, some WD docs show DMA mw2, some PIO4...), kernel 2.2.17 with ide patch and PIIX enabled, DMA by default, generic DMA, and couple others. Did not try 'bad DMA-firmware (EXPERIMENTAL)'. dmesg (relevant - in my view - lines):

 Linux version 2.2.17 (root@risky) (gcc version 2.7.2.3) #22 Wed Nov 29 23:04:12
 GMT 2000
 Detected 74540 kHz processor.
 ...
 Memory: 47168k/49152k available (828k kernel code, 412k reserved, 700k data,
 44k init)
 ...
 CPU: Intel Pentium 75 - 200 stepping 05
 ...
 PCI: PCI BIOS revision 2.10 entry at 0xfcb81
 PCI: Using configuration type 1
 PCI: Probing PCI hardware
 ...
 Uniform Multi-Platform E-IDE driver Revision: 6.30
 ide: Assuming 33MHz system bus speed for PIO modes; override with idebus=xx
 PIIX: IDE controller on PCI bus 00 dev 38
 PIIX: chipset revision 2
 PIIX: not 100% native mode: will probe irqs later
 PIIX: neither IDE port enabled (BIOS)
 hda: WDC AC21600H, ATA DISK drive
 hdc: BCD-16X 1997-03-25, ATAPI CDROM drive
 ide0 at 0x1f0-0x1f7,0x3f6 on irq 14
 ide1 at 0x170-0x177,0x376 on irq 15
 hda: WDC AC21600H, 1549MB w/128kB Cache, CHS=787/64/63
 ...
 Partition check:
  hda: hda1 hda2 hda3
 VFS: Mounted root (ext2 filesystem) readonly.
 ...

BIOS DOES identify the hard disk and the CD-ROM correctly, although it is pretty old and no newer version is available. hdparm -d1 returns:
 /dev/hda:
  setting using_dma to 1 (on)
  HDIO_SET_DMA failed: Operation not permitted
  using_dma    =  0 (off)
 hdparm -I:
 /dev/hda:

  Model=DW CCA1206H0                            , FwRev=420.P980,
 SerialNo=DWW-3M63
  Config={ HardSect NotMFM HdSw>15uSec SpinMotCtl Fixed DTR>5Mbs FmtGapReq }
  RawCHS=3148/16/63, TrkSize=57600, SectSize=600, ECCbytes=22
  BuffType=3(DualPortCache), BuffSize=128kB, MaxMultSect=16, MultSect=16
  DblWordIO=no, maxPIO=2(fast), DMA=yes, maxDMA=0(slow)
  CurCHS=3148/16/63, CurSects=3173184, LBA=yes, LBAsects=3173184
  DblWordIO=no, maxPIO=2(fast), DMA=yes, maxDMA=0(slow)
  CurCHS=3148/16/63, CurSects=3173184, LBA=yes, LBAsects=3173184
  tDMA={min:120,rec:120}, DMA modes: mword0 mword1 *mword2
  IORDY=on/off, tPIO={min:160,w/IORDY:120}, PIO modes: mode3 mode4

(yesterday I installed hdparm-3.9, now output is in slightly different format, but the same contents). hdparm -v reports dma off. lspci looks fine... hdparm -tT gives 4.8MB/s. I just downloaded a patch by Andre Hedrick for hdparm-3.9a (why 'a', BTW? is it the patch to hdparm-3.9 actually?). Will try it tonight - would that help? I'll rather stop trying to foresee what information is needed here, it is already TOO long, sorry.

Thanks in advance
Guennadi
-
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to majordomo@vger.kernel.org
Please read the FAQ at http://www.tux.org/lkml/

Re: DMA for triton again...

[Mike Dresser]

> computer: Chipset 430FX / Triton / PIIX, disk Western Digital Caviar AC21600H Firmware code F6 (no UDMA, some WD docs show DMA mw2, some PIO4...), kernel 2.2.17 with ide patch and PIIX enabled, DMA by default, generic DMA, and couple others. Did not try 'bad DMA-firmware (EXPERIMENTAL)'. dmesg (relevant - in my view - lines):
> BIOS DOES identify the hard disk and the CD-ROM correctly, although it is pretty old and no newer version is available.

Running a basically similar system here, DataExpert 8551, with a modified Award (was originally an ami) bios from http://exp8551.mypage.org.
Chipset is a 430FX, Same hard drive as what you have.  Pentium 133, 48 meg ram.  Kernel 2.2.17 with raid patch, and ide patch.  Shares the ide cable with a WDC 850 meg drive as slave.

monitor:~# hdparm -d1 /dev/hda

/dev/hda:
 setting using_dma to 1 (on)
 using_dma    =  1 (on)

> hdparm -d1 returns:
>  /dev/hda:
>   setting using_dma to 1 (on)
>   HDIO_SET_DMA failed: Operation not permitted
>   using_dma    =  0 (off)
>

/dev/hda:

 Model=WDC AC21600H, FwRev=24.09P07, SerialNo=WD-WM3362938634
 Config={ HardSect NotMFM HdSw>15uSec SpinMotCtl Fixed DTR>5Mbs FmtGapReq }
 RawCHS=3148/16/63, TrkSize=57600, SectSize=600, ECCbytes=22
 BuffType=DualPortCache, BuffSize=128kB, MaxMultSect=16, MultSect=off
 CurCHS=3148/16/63, CurSects=3173184, LBA=yes, LBAsects=3173184
 IORDY=on/off, tPIO={min:160,w/IORDY:120}, tDMA={min:120,rec:120}
 PIO modes: pio0 pio1 pio2 pio3 pio4
 DMA modes: mdma0 mdma1 *mdma2

so my firmware is actually one version older than yours, and allows me to set DMA.  And has the associated speed increase/lower cpu

>   Model=DW CCA1206H0                            , FwRev=420.P980,
>  SerialNo=DWW-3M63

What brand is your motherboard?

Mike Dresser

-
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to majordomo@vger.kernel.org
Please read the FAQ at http://www.tux.org/lkml/

Re[2]: DMA for triton again...

[Guennadi Liakhovetski]

Thanks!

> Chipset is a 430FX, Same hard drive as what you have.  Pentium > 133, 48 meg ram.  Kernel 2.2.17 with raid patch, and ide patch.  

I don't think I need tha raid patch, do I?

> Shares the ide cable with a WDC 850 meg drive as slave.

No slaves on mine.

> What brand is your motherboard?

Intel Morrison64 (not Morrison32!) aka Advanced/MN (in some sources Advanced/AL). BIOS AMI 1.00.03.CA0 (upgraded to 1.00.04.CA0).

What version of hdparm are you using? By the output it looks like 3.9, patched? Did DMA work from the very beginning? Can you send me a copy of your .config file?

Regards
Guennadi
-
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to majordomo@vger.kernel.org
Please read the FAQ at http://www.tux.org/lkml/

Re: DMA for triton again...

[Mike Dresser]

Guennadi Liakhovetski wrote:

> Thanks!
>
> > Chipset is a 430FX, Same hard drive as what you have.  Pentium > 133, 48 meg ram.  Kernel 2.2.17 with raid patch, and ide patch.
>
> I don't think I need tha raid patch, do I?

No, its just left overs from my playing with a Promise Ultra/66 card.  Even without the raid and ide patches, i can use DMA

> Intel Morrison64 (not Morrison32!) aka Advanced/MN (in some sources Advanced/AL). BIOS AMI 1.00.03.CA0 (upgraded to 1.00.04.CA0).
>

Opposite side of the spectrum from the utter crap this DataExpert is. =)  Had to get a modified bios, to get a 13.6 gig to exist in the machine without locking it up
on the POST.  Yes, you can tell it no bios, but without the ide patches, it's _very_ slow.

> What version of hdparm are you using? By the output it looks like 3.9, patched? Did DMA work from the very beginning? Can you send me a copy of your .config file?

standard hdparm 3.9, from debian woody.  The difference is i used -i, instead of -I, i think.  I'll see how mangled my .config is.
Looks somewhat clean, a few modules i don't use.  I'll send it in a seperate email.

I've got maybe 3 or 4 of these machines in service as backup servers for small lans.  Small 1.6 gig boot drive, 13.6 gig data drive, a cdrom, and a IDE 7/14 gig HP
tape drive.  Speeds aren't too bad, and DMA seems to work on the parts that support it.


-
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to majordomo@vger.kernel.org
Please read the FAQ at http://www.tux.org/lkml/

Re[2]: DMA for triton again...

[Guennadi Liakhovetski]

> I'll see how mangled my .config is.
> Looks somewhat clean, a few modules i don't use.  I'll send it in a > seperate email.

Thanks a lot, Mike! I can see couple of differences already, will try tonight... and will let you know tomorrow...

Cheers
Guennadi

-
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to majordomo@vger.kernel.org
Please read the FAQ at http://www.tux.org/lkml/

Re: DMA for triton again...

[Mike Dresser]

    You're going to get a chuckle out of this one, i think.  I was using my main fileserver at home(different box from
the ones we're talking about)last night, and figured i'd poke around, do some speed benchmarks for the hell of it.  At
one point months, i had everything running in DMA mode, drives are doing 20ish meg a second transfers.  Had to replace
my EDO 128 meg dimm with a 32 meg SDRAM stick, because quite frankly, EDO dimms suck at 100fsb. :P  So anyways, i
check, and lo and behold, my system is running all the drives in non-DMA mode.  THE HORROR!  I was wondering why the
network sucked :)  So i do some poking around.. Dumped the IDE patches, didn't make a difference.  Put the ide patches
back in, started playing in the bios.  Finally fixed it.  I was getting the same operation not permitted, that you
were,until i got that bios setting.  My mission tonight will be to find which one it was =)  The weird part is that
this is a 591 Apollo Pro chipset, not the FX.  But it's making me wonder if it's something similar in your bios!
I know it wasn't the actual UDMA setting in the bios, i'm wondering what it was though.  I'll put a keyboard on it,
and poke around tonight or this weekend.


Guennadi Liakhovetski wrote:

> > I'll see how mangled my .config is.
> > Looks somewhat clean, a few modules i don't use.  I'll send it in a > seperate email.
>
> Thanks a lot, Mike! I can see couple of differences already, will try tonight... and will let you know tomorrow...
>
> Cheers
> Guennadi
>
> -
> To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
> the body of a message to majordomo@vger.kernel.org
> Please read the FAQ at http://www.tux.org/lkml/

-
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to majordomo@vger.kernel.org
Please read the FAQ at http://www.tux.org/lkml/

Re[3]: DMA for triton again...

[Guennadi Liakhovetski]

Hi
> will try tonight... and will let you know tomorrow...
... Nop, it didn't work. Mike and everybody having experience / knowledge of Western Digital Caviar AC21600H... If you compare WD's documents at:
http://www.westerndigital.com/service/FAQ/dtr.html
and
http://www.westerndigital.com/products/non-current/drives/ac21600.html
they, respectively ,say:

UDMA if CCC is A8-AG (AH is not UDMA), D5-D7, Bx, or Cx All others -- PIO Mode 4

and

16.6 MB/s (burst Mode 4 PIO) *
16.6 MB/s (burst Mode 2 DMA) **
33.3 MB/s (Mode 2 Ultra DMA) ***

*Max PIO burst rate is specified at 16.6 MB/s using the IORDY signal. 
**Max DMA and multi-word DMA burst rate is specified using the DMARQ and DMACK signals.
***Mode 2 Ultra DMA is supported in the following firmware revisions (CCC Codes: A8, AA, AC, AD, AG, D5, D6, D7, Bx, Cx). 

Which one is true? Do non-UDMA AC21600H (I've got CCC F6) support DMA?

Is it possible to find out where hdparm -d1 is stopped? Maybe by compiling it with -g and debugging... As far as I understand, BIOS is not an issue, since ide.txt says, that Linux doesn't use BIOS when working with hard drives...

Thanks
Guennadi
-
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to majordomo@vger.kernel.org
Please read the FAQ at http://www.tux.org/lkml/

Re: DMA for triton again...

[Mike Dresser]

Guennadi Liakhovetski wrote:

> Hi
> > will try tonight... and will let you know tomorrow...
> ... Nop, it didn't work. Mike and everybody having experience / knowledge of Western Digital Caviar AC21600H... If you compare WD's documents at:
> http://www.westerndigital.com/service/FAQ/dtr.html
> and
> http://www.westerndigital.com/products/non-current/drives/ac21600.html
> they, respectively ,say:
>
> UDMA if CCC is A8-AG (AH is not UDMA), D5-D7, Bx, or Cx All others -- PIO Mode 4
>
> and
>
> 16.6 MB/s (burst Mode 4 PIO) *
> 16.6 MB/s (burst Mode 2 DMA) **
> 33.3 MB/s (Mode 2 Ultra DMA) ***
>
> *Max PIO burst rate is specified at 16.6 MB/s using the IORDY signal.
> **Max DMA and multi-word DMA burst rate is specified using the DMARQ and DMACK signals.
> ***Mode 2 Ultra DMA is supported in the following firmware revisions (CCC Codes: A8, AA, AC, AD, AG, D5, D6, D7, Bx, Cx).
>
> Which one is true? Do non-UDMA AC21600H (I've got CCC F6) support DMA?
>
> Is it possible to find out where hdparm -d1 is stopped? Maybe by compiling it with -g and debugging... As far as I understand, BIOS is not an issue, since ide.txt says, that Linux doesn't use BIOS when working with hard drives...
>
> Thanks
> Guennadi
> -
> To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
> the body of a message to majordomo@vger.kernel.org
> Please read the FAQ at http://www.tux.org/lkml/

Well, per my other email, i found a case where the bios DID matter.  Yes, the AC21600H supports DMA.  Even my dinosaur 850 meg in that same box, works in dma mode, and does get a little bit faster.  They're so slow in the first place
that it doesn't make much transfer rate difference, but the cpu load goes way down.



-
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to majordomo@vger.kernel.org
Please read the FAQ at http://www.tux.org/lkml/